Pattern distillation methods in grammar induction

The natural language processing module is an important implementation form of intelligent human-machine interaction. The NLP engine requires an efficient grammar description to convert sentences into semantic models. Automated grammar induction is the area to discover rules in the sentences of a given language. An important phase of grammar induction is the pattern distillation phase. The paper presents a survey of pattern distillation methods for generation of context free grammars. The paper proposes a link-based pattern distillation method as a good alternative of the most widely used MEX algorithm. Beside the methodological analysis, test results are also presented to compare the efficiency of the main pattern distillation algorithms.